Let's keep the atmosphere, at the very least so we can continue breathing ;)
What this boils down to is whether there's more delta-V from drag during launch vs. delta-V provided by aerobraking during re-entry. You could argue about the costs associated with heat-resistant panels, but most of your rocket-dollars are going to be put towards other parts of the system.
In a few cases, like with the ISS, you have to consider long-term atmospheric drag, but most objects are in high enough orbits that they're not appreciably losing velocity to drag - after all, we don't re-boost most of them. Satellites will de-orbit after a long enough period of time, but I choose to see that as a perk, given the alternative would be extensive debris junking up the orbital slots we prefer.
Wikipedia tells us we need about 9.4 km/s to reach LEO with an orbital speed of 7.8 km/s. The same section tells us atmospheric and gravity account for the difference, though it ranges between 1.3 and 1.8 km/s. As Russell Borogove points out, most of that is due to gravity and the atmospheric value tops out around a couple hundred
When something comes back into the atmosphere from LEO, it's going to have at least 7.8 km/s of re-entry dV. The vast majority of that is diminished by aerobraking, and the remainder dissipated by chutes, gliding or boosters, depending on the vehicle.
TL;DR: A couple hundred m/s << 7.8 km/s. Atmospheres are helpful
Edit: Changed atmospheric contribution numbers per Russell Borogove's comment